Background: Terminal digit and single-number preference may produce inaccuracy and biased results when measuring blood pressure. We describe these preferences in the Syst-Eur randomized placebo-controlled trial of the treatment of isolated systolic hypertension and describe how we sought to eliminate these problems.
Methods: The Data Monitoring Committee of the trial conducted yearly quality control meetings in Belgium and visited the participating centres to check their adherence to the protocol. These meetings involved identifying terminal digit preference, improving blood pressure control and boosting recruitment.
Results: The prevalence of use of terminal digit zero when measuring sitting systolic blood pressure (first readings) reduced from an average of 42.4% in the year prior to the date when a centre first randomized a patient to 31.5, 25, 22.3, 26.3, 23.2 and 22% in the subsequent 6 years. This trend was independent of the calendar year during which a centre entered the trial and supports the hypothesis that data-quality monitoring, including the feedback of digit preference to centres, led to a reduction in terminal digit zero preference. In addition, a higher than expected prevalence of the systolic blood pressure value of 148 mmHg was found in the active treatment groups in the double-blind phase. Selection for 148 mmHg persisted over time and constituted a single-number preference bias. This arose from the instruction to investigators to reduce systolic blood pressure to below 150 mmHg.
Conclusion: Monitoring and feedback of data quality should be undertaken to minimize digit and number preference. Automatic devices should ideally be employed to help to avoid these problems as long as the devices are fully validated and regularly serviced, and providing that readings are not rejected and repeated.